This paper presents a new method for visual correspondence that properly handles occlusions while maintaining the advantages of graph cut algorithms. The method is based on energy minimization and ensures that each pixel corresponds to at most one pixel in the other image, enforcing uniqueness. The authors define a problem representation and an energy function that penalizes non-unique configurations and uses graph cuts to find a strong local minimum of this energy function. Experimental results on stereo and motion data demonstrate the effectiveness of the method in detecting occlusions and computing disparities. The algorithm is compared with existing methods, showing superior performance in handling occlusions and discontinuities. The paper also discusses related work and provides a detailed analysis of the energy function and graph construction.This paper presents a new method for visual correspondence that properly handles occlusions while maintaining the advantages of graph cut algorithms. The method is based on energy minimization and ensures that each pixel corresponds to at most one pixel in the other image, enforcing uniqueness. The authors define a problem representation and an energy function that penalizes non-unique configurations and uses graph cuts to find a strong local minimum of this energy function. Experimental results on stereo and motion data demonstrate the effectiveness of the method in detecting occlusions and computing disparities. The algorithm is compared with existing methods, showing superior performance in handling occlusions and discontinuities. The paper also discusses related work and provides a detailed analysis of the energy function and graph construction.